Just an update on this one, the latest code has improved things a little in that it no longer displays green corruption. I think this is due to this commit:
However, this still means that on key frames it is restarting the decoder due to the issue, and according to the trace is dropping a frame each time (picture too late to be displayed).
So the underlying issue still needs solving, but the effect isn't as bad.
If you want to create a new bug report, that's fine.
Personally I can't really see the point, as you'd want to link back to this one in any case, or transfer the attachments and discussion, as it's the same underlying cause regardless of the effect on the screen.
Well it's the same bug with the same cause as discussed on the mailing list and linked to above. i.e. the decoder is getting reset apparently spuriously (I say spuriously because when I hacked it to not reset the decoder with his PPS change as I discussed on the list, it fixes the issue at least for my test file)...It's just a different effect now.
As long as you link back to this bug then I guess there is no problem with opening a new bug report!